Design: systematic review and meta-analysis of cross-sectional studies
Population: 1359 Indian adults with type 2 diabetes across seven studies
Primary outcome: pooled prevalence of confirmed sarcopenia by AWGS 2019 criteria
Effect: 27.4% (95% CI 18.9-37.9), I2 = 91.7%, prediction interval 6.9-65.6%

Seven cross-sectional Indian studies, 1359 adults with type 2 diabetes in total, were pooled using the Asian Working Group for Sarcopenia criteria. Confirmed sarcopenia came to 27.4% (95% CI 18.9% to 37.9%), severe sarcopenia to 11.0%, and sarcopenic obesity to 25.2%.

The pooled figure is fragile. Heterogeneity was 91.7% and the prediction interval - the range a new cohort would plausibly fall in - ran from 6.9% to 65.6%. Measurement method explained part of it: 20.2% by DXA against 31.7% by bioelectrical impedance. GRADE certainty was very low for every prevalence outcome, and the authors explicitly decline to recommend screening on this basis.

What survives is the clinical awareness. Indian body composition at a given BMI carries less muscle, the AWGS cut-offs are the right ones to use here rather than the European criteria, and a patient losing weight on an incretin without resistance training is losing some of that muscle. Grip strength and a five-times sit-to-stand cost nothing to check.

Use the Asian Working Group cut-offs, not European ones, when assessing muscle mass in Indian patients.
Grip strength and five-times sit-to-stand need no equipment and identify possible sarcopenia in clinic.
Bioelectrical impedance reads sarcopenia commoner than DXA - know which your report used.
Check resistance training and protein intake in anyone losing weight on incretin therapy.
Hypertension and diabetes duration of 10 years or more were associated with sarcopenia, though imprecisely.

Why it matters

Muscle loss is the side of Indian diabetes care that the BMI-centred consultation misses entirely.

The statistics, in plain English

An I2 of 91.7% means almost all the variation between these seven studies is real disagreement rather than chance, so the single pooled number of 27.4% describes none of them well. The prediction interval is the honest summary: anywhere from 7% to 66%. That is why the authors treat this as a call for better studies rather than an estimate to act on.